Massachusetts Institute of Technology
Department of Urban Studies and Planning
11.520: A Workshop on Geographic Information
Systems
11.188: Urban Planning and Social Science Laboratory
Lecture 4: Relational Databases & GIS Data Models
September 28, 2005, Joseph Ferreira, Jr.
(including contributions from Visting Prof. Zhong-Ren Peng who taught the class
Fall, 2003)
GIS Data Models
„
Computer Aided Design (CAD), graphical, and
image GIS data models;
„
Raster data model;
„
Vector data model;
„
Linear Reference data model;
„
Network data model;
„
TIN data model.
The contents of the GIS data model section of today’s lecture is
derived from Longley, Goodchild, Maguire and Rhind, Geographic
Information Systems and Science, 2001, as organized by Prof.
Zhong-Ren Peng for 11.520 in Fall 2003.
CAD data models
„
„
In CAD, real-world entities are represented
symbolically as points, lines and polygons.
CAD data model is different from GIS data
models:
„
„
„
CAD models uses local drawing coordinates rather
than real-world coordinates.
Individual objects in CAD do not have unique
identifiers and attributes.
CAD data model does not store details of
relationships (e.g., topology) between objects.
Computer Cartography
„
„
„
„
The purpose of computer cartography is to
automatically reproduce paper maps.
All paper map entities are stored as points, lines
and polygons, with annotations used for
placement.
Only limited attribute data are associated with
entities (enough for symbology).
Relationships among entities are not stored.
Image data model
„
„
Use images (photos, aerial photos and satellite
images) to represent real world entities.
Working with annotated pictures or real world
entities.
„
Images need rectification and registration to be
integrated with other georeference data. (e.g.,
GeoTiff vs. Tiff image format)
Raster data model
„
„
„
„
The raster data model uses an array of cells, or
pixels, to represent real-world space and then
encodes the grid cells based on objects in the cell.
The cells can hold any attribute values based on
different encoding schemes.
Raster data are usually stored as an array of grid
values, with metadata held in a file header.
Typical metadata includes geographic coordinate
of the upper-left corner of the grid, the cell size,
and the number of row and column elements.
An example of a raster dataset with integer and
floating point grid cell values (e.g., layer1 = soil
type at center of grid cell, layer2 = cell average
groundwater depth in meters)
Row Column Layer1_value Layer2_value
1
1
1
2.0090
1
2
0
12.665
1
3
2
1.2211
2
1
1
3.3566
...
...
Vector Data Model
„
„
„
Vector data model describes the boundaries of
real-world objects using 2D geometric types: point,
line, or polygon and a real-world coordinate
system
Simple features model the basic object geometry
Topologic features model relationsihps among
objects
Simple Features
„
„
„
„
„
„
Geographic entities encoded using the vector data
model are called features.
Features are vector objects of type point, line or
polygon.
Lines and polygons can overlap.
There is no stored relationship between any
objects.
Simple feature data structure is sometimes called
spaghetti.
See textbook Longley et al. (2001) pp. 190.
Advantages and drawbacks of simple features
„
Easy to create and store.
„
Easy to retrieve and render on screen.
„
„
„
„
Inefficient to store, boundaries of two adjacent
polygons - shared boundaries need to be stored
twice.
Inflexible in dissolving common boundaries when
joining two zones or editing geometry to move
common boundary points.
May result in gaps (slivers) or overlaps of
polygons.
Many operations cannot be performed due to the
lack of connectivity relationships in the data
structure: e.g., finding the shortest path through a
road network.
Topologic features
„
„
„
„
Topologic features are simple features structured
using topologic rules.
Topology is the science and mathematics of
encoding shape and spatial relationships.
Topology can be used to validate the geometry of
vector entities (e.g., polygons that aren't 'closed').
Topology can be used for certain operations such
as network tracing and tests for polygon
adjacency.
Topologic structure – Line
„
A line is defined as a directed sequence of points
from a starting node to an ending node.
„
„
„
Points (vertices or line end nodes) that fall within
a minimum tolerance of each other are snapped
together.
New nodes are created wherever two lines
intersect.
Hence, the name “spaghetti with meatballs".
Topologic structure – polygon
„
„
„
Polygons are defined as a sequence of lines that
enclose an area.
Points are stored once in a point list; lines are
stored as a directed sequence of point IDs;
polygons are stored as a directed sequence of line
IDs. Moving a point, will automatically move all
the lines and polygons that utilize the point.
<![endif]> Planar
enforcement: all the space on a
map must be filled and any point must fall in one
polygon only. That is, polygons must not overlap.
PointID
X-value
Y-value
1
200101.22
620112.3
2
150000.00
510000.4
3
300100.10
777000.1
4
300000.00
400000.00
5
260000.00
200000.00
LineID
FromNode
ToNode
10
2
1
20
2
5
30
5
4
40
4
1
50
1
3
60
4
3
PolygonID
LineID sequence
A
20, 10, 40, 30
B
50, 60, 40
Contiguity (adjacency) relationship
„
„
Each line must have a direction to define
contiguity (adjacency) relationship.
Defined by a list of polygons on the left- and
right-hand side of each line.
„
An example is given at the textbook Longley et al
(2001) pp. 191, Figure 9.9.
Geo-Relation model
„
„
„
„
The Geo-relation model refers to software that
implements the vector topologic feature data
model.
Store the geometry and associated topologic
information in one set of files (with object IDs).
Store associated attribute information in relational
database management system (with object IDs as
foreign keys into geometry and topology data).
The GIS software maintains the linkage between
geometry, topology and attribute information.
Network data model
„
„
„
„
Networks are modeled as points (nodes) and lines.
Network topology defines how lines connect with
each other at nodes.
There are also rules to define how flows can move
through a network, like one-way street, sewerage
flow, etc.
The rate of flow is modeled as impedance on the
nodes and lines, such as turn signals and speed
limits.
Linear Referencing
„
„
„
Store the location of geographic entities (called
events) as a distance along a network (a route
system) from a point of origin.
Offsets are used to store information about the
distance from the centerline.
Dynamic segmentation is a special type of linear
referencing. It assign event data to the network
segments dynamically.
Triangulated Irregular Network (TIN) data
model
„
„
„
„
TINs are used to create and represent surfaces in
GIS.
The TIN structure represents a surface as
contiguous non-overlapping triangular elements.
TIN is created from a set of sample points with x,
y, and z coordinate values.
An example of TIN data sturcture is given at the
textbook Longley et al (2001) pp. 195, figures
9.12 and 9.13.
Data Model Examples from Lab Exercises
o
Vector: objects as points, lines, polygons
ƒ ArcGIS data files:
ƒ Coverages: old Arc/Info a directory per
layer, plus INFO files
Shapefiles: .shp, .shx, .dbf files (and
possibly others)
ƒ Spatial Database Engine (SDE):
retrieved dynamically from a database
server
ƒ Lab exercises 1 and 2:
ƒ Cambridge blockgroups (polygons in
shapefile)
ƒ Look at cambbgrp.dbf from MSAccess (beware of sort order for
matching geometry)
ƒ Sales89 (points in shapefile)
ƒ Can't view sales89_shape.dbf in
MS-Access because name is too
long for Access!
ƒ Lat/lon columns in a data table can
be used by ArcMAP to create
point geometry and build a
shapefile (the sales89 shapefile
was built by geocoding street
address to lat/lon and building
point layer from the lat/lon)
Raster: space as 2D or 3D grid cells
ƒ regular grid on top of spatial features
(instead of encoding boundary)
ƒ pixel brightness in orthophoto of Boston
ƒ Raster example: orthophotos, scanned maps,
grids
ƒ Lab exercise 2:
ƒ MITOrthoTool brings in orthophoto
snippets as TIFF images
ƒ
o
ƒ
TIFF image doesn't know 'where' it is
but ArcMAP script has already figured
that out
Boston/Cambridge Streets superimposed on orthophoto. Zoomed-in view shows
raster nature of the ortho.
•
Complications
o islands, lakes, overpasses
o share edges?, move links when you move points?
o ambiguity: summer/winter wetland boundaries
o
scale/level-of-detail, generalization, conflation,
slivers
Relational Databases
•
•
•
One 'flat file' data table for each map layer is too
simple a data model
o From lab exercise #3: because cambbgrp was a
read-only shapefile, we had to build a new table
with percent-with-high-school-education
computations and then join it to the original
cambbgrp table
o From lab exercise #3: some houses in sales89
may have sold more than once! But, there's only
one dot for the location of the house
o Keep tabular model for handling textual data but
allow many tables that can be related to one
another via common columns (attributes)
ƒ Allows us to extend attribute tables by
adding additional data for each blockgroup
ƒ Handles one-to-many situations such as
multiple sales for the same house
ƒ Provides many other powerful mix-andmatch capabilities
The relational database model and the structured query
language (SQL) for joining tables and specifying
queries is the lingua franca of distributed database
operations on public (internet) and private networks
(bank ATM transactions, airline reservations)
Continue in Monday's lab and next Wednesday's
lecture with more on relational database management
and queries - especially, handling one-to-many
relationships via ArcMap 'summarize' tools, and MSAccess queries.
Last modified 28 September 2005. [jf]